[转载]关于讯飞语音SDK开发学习 – 或、许 – 博客园.
前奏,浑浑噩噩已经工作一年多,这一年多收获还是挺多的。逛园子应该有两年多了,工作后基本上是天天都会来园子逛逛,园子 里还是有很多牛人写了一些不错的博客,帮我解决很多问题。但是一直没写过博客,归根到底一个字“懒”,还有就是不知道该写 些什么…
今 天把我最近研究讯飞语音东东,分享一下,不过有些还是前辈们提供的。之前公司让我做一个小的语音识别功能,一开始我就建议使用讯飞语音,个人觉得讯飞识别 正确率还是可观的。可是老总说不能考 虑联网,还有就是钱的问题。想到微软自带语音识别引擎(基于win7)。第一次接触到语音识别,没什么头绪,只有收集相关资料 。最后成品出来了,但是识别效果不是那么满意,老总说那就将就用吧,我想他都那样说了,我也没多大意见…开年后老总老总 买了个iphone5玩了siri后,觉得我们现在那个语音太丑陋了,让换一个解决方案,识别率要达到90%。国内有好几家公司做语音识 别的比如科大讯飞、云知声、捷通华声以及紫冬锐意语音都做了一定开放。市面上我知道语音助手有百度语音助手、虫洞语音助手 、360语音助手以及科大讯飞语点;前两个我不知道采用那家公司的还是自己研发的不过都没开开放,360语音助手采用讯飞的。然 后我就继续提议使用讯飞语音SDK,于是乎同事们都下载讯飞语点来玩玩,老总说那就试试讯飞语音SDK。
先 做一个简单demo,看看识别效果,感觉识别率上能够满足要求。一般要的结果不光只是要把所说的话翻译成文字,而是需要的是 语义的理解:例如我要去北京,直接返回北京这个关键。目前讯飞还没把这个接口开放出来,公司负责人说今年会把这个开放出来 。那现在只能使用关键词识别语法,一种是直接是文本词库;另一种是ABNF语法。ABNF写法有点烦杂(语法文件里使用到的词句都是指定的,对于无法枚举 的词句暂时没有很好的解决办法。),就直接采用文本词库(因为关键词有点多)。讯飞还有专门人负责关于讯飞语音问题解答(QQ群:153789256)。
讯飞提供msc.dll这个DLL,调用DLL的封装:
/// <summary> /// MSCDLL入口封装 /// </summary> private class MscDll { #region MscDLL /// <summary> /// 初始化MSC的ISR部分 /// </summary> /// <param name="configs">初始化时传入的字符串,以指定合成用到的一些配置参数,各个参数以“参数名=参数值”的形式出现,大小写不敏感,不同的参数之间以“,”或“\n”隔开,不设置任何值时可以传入NULL或空串:</param> /// <returns>如果函数调用成功返回MSP_SUCCESS,否则返回错误代码,错误代码参见msp_errors</returns> [DllImport("msc.dll", CallingConvention = CallingConvention.StdCall)] public static extern int QISRInit(string configs); /// <summary> /// 开始一个ISR会话 /// </summary> /// <param name="grammarList">uri-list格式的语法,可以是一个语法文件的URL或者一个引擎内置语法列表。可以同时指定多个语法,不同的语法之间以“,” /// 隔开。进行语音听写时不需要语法,此参数设定为NULL或空串即可;进行语音识别时则需要语法,语法可以在此参数中指定,也可以随后调用QISRGrammarActivate指定识别所用的语法。</param> /// <param name="_params">本路ISR会话使用的参数,可设置的参数及其取值范围请参考《可设置参数列表_MSP20.xls》,各个参数以“参数名=参数值” 的形式出现,不同的参数之间以“,”或者“\n”隔开。</param> /// <param name="errorCode">如果函数调用成功则其值为MSP_SUCCESS,否则返回错误代码,错误代码参见msp_errors。几个主要的返回值: MSP_ERROR_NOT_INIT 未初始化 MSP_ERROR_INVALID_PARA 无效的参数; MSP_ERROR_NO_LICENSE 开始一路会话失败</param> /// <returns>MSC为本路会话建立的ID,用来唯一的标识本路会话,供以后调用其他函数时使用。函数调用失败则会返回NULL。</returns> [DllImport("msc.dll", CallingConvention = CallingConvention.StdCall)] public static extern IntPtr QISRSessionBegin(string grammarList, string _params, ref int errorCode); /// <summary> /// 传入语法 /// </summary> /// <param name="sessionID">由QISRSessionBegin返回过来的会话ID</param> /// <param name="grammar">语法字符串</param> /// <param name="type">语法类型,可以是uri-list、abnf、xml等</param> /// <param name="weight">本次传入语法的权重,本参数在MSP 2.0中会被忽略。</param> /// <returns>如果函数调用成功返回MSP_SUCCESS,否则返回错误代码,错误代码参见msp_errors</returns> [DllImport("msc.dll", CallingConvention = CallingConvention.StdCall)] public static extern int QISRGrammarActivate(string sessionID, string grammar, string type, int weight); /// <summary> /// 写入用来识别的语音 /// </summary> /// <param name="sessionID">由QISRSessionBegin返回过来的会话ID</param> /// <param name="waveData">音频数据缓冲区起始地址</param> /// <param name="waveLen">音频数据长度,其大小不能超过设定的max_audio_size</param> /// <param name="audioStatus">用来指明用户本次识别的音频是否发送完毕,可能值如下: /// MSP_AUDIO_SAMPLE_FIRST = 1 第一块音频 /// MSP_AUDIO_SAMPLE_CONTINUE = 2 还有后继音频 /// MSP_AUDIO_SAMPLE_LAST = 4 最后一块音频</param> /// <param name="epStatus">端点检测(End-point detected)器所处的状态,可能的值如下: /// MSP _EP_LOOKING_FOR_SPEECH = 0 还没有检测到音频的前端点。 /// MSP _EP_IN_SPEECH = 1 已经检测到了音频前端点,正在进行正常的音频处理。 /// MSP _EP_AFTER_SPEECH = 3 检测到音频的后端点,后继的音频会被MSC忽略。 /// MSP _EP_TIMEOUT = 4 超时。 /// MSP _EP_ERROR= 5 出现错误。 /// MSP _EP_MAX_SPEECH = 6 音频过大。</param> /// <param name="recogStatus">识别器所处的状态</param> /// <returns></returns> [DllImport("msc.dll", CallingConvention = CallingConvention.StdCall)] public static extern int QISRAudioWrite(string sessionID, byte[] waveData, uint waveLen, AudioStatus audioStatus, ref EpStatus epStatus, ref RecogStatus recogStatus); /// <summary> /// 获取识别结果 /// </summary> /// <param name="sessionID">由QISRSessionBegin返回过来的会话ID</param> /// <param name="rsltStatus">识别结果的状态,其取值范围和含义请参考QISRAudioWrite的参数recogStatus</param> /// <param name="waitTime">与服务器交互的间隔时间,可以控制和服务器的交互频度。单位为ms,建议取值为5000</param> /// <param name="errorCode">如果函数调用成功返回MSP_SUCCESS,否则返回错误代码,错误代码参见msp_errors</param> /// <returns>函数执行成功并且获取到识别结果时返回识别结果,函数执行成功没有获取到识别结果时返回NULL</returns> [DllImport("msc.dll", CallingConvention = CallingConvention.StdCall)] public static extern IntPtr QISRGetResult(string sessionID, ref RecogStatus rsltStatus, int waitTime, ref int errorCode); /// <summary> /// 结束一路会话 /// </summary> /// <param name="sessionID">由QISRSessionBegin返回过来的会话ID</param> /// <param name="hints">结束本次会话的原因描述,用于记录日志,便于用户查阅或者跟踪某些问题</param> /// <returns></returns> [DllImport("msc.dll", CallingConvention = CallingConvention.StdCall)] public static extern int QISRSessionEnd(string sessionID, string hints); /// <summary> /// 获取与识别交互相关的参数 /// </summary> /// <param name="sessionID">由QISRSessionBegin返回过来的会话ID</param> /// <param name="paramName">要获取的参数名称;支持同时查询多个参数,查询多个参数时,参数名称按“,” 或“\n”分隔开来</param> /// <param name="paramValue">获取的参数值,以字符串形式返回;查询多个参数时,参数值之间以“;”分开,不支持的参数将返回空的值</param> /// <param name="valueLen">参数值的长度</param> /// <returns></returns> [DllImport("msc.dll", CallingConvention = CallingConvention.StdCall)] public static extern int QISRGetParam(string sessionID, string paramName, string paramValue, ref uint valueLen); /// <summary> /// 逆初始化MSC的ISR部分 /// </summary> /// <returns></returns> [DllImport("msc.dll", CallingConvention = CallingConvention.StdCall)] public static extern int QISRFini(); /// <summary> /// 上传用户自定义词库 /// </summary> /// <param name="sessionID">由QISRSessionBegin返回过来的会话ID</param> /// <param name="dataName">词库名称</param> /// <param name="userData">词库数据采用是utf8格式</param> /// <param name="lenght">词库大小</param> /// <param name="paramValue">参数值</param> /// <param name="errorCode">如果函数调用成功返回MSP_SUCCESS,否则返回错误代码,错误代码参见msp_errors</param> /// <returns>函数执行成功并且返回exID(词库编号)</returns> [DllImport("msc.dll", CallingConvention = CallingConvention.StdCall)] public static extern IntPtr QISRUploadData(string sessionID, string dataName, byte[] userData, uint lenght, string paramValue, ref int errorCode); #endregion }
说明一下:“QISRUploadData”(上传词库)这个函数在开发文档里面没的,讯飞遗漏了。
具体实现(关键词识别,文本词库):
类:MscNet
#region 定义字段 //返回错误代号 private int ret = 0; private RecoErrors Re = null; /// <summary> /// 会话ID /// </summary> private string sess_id = null; /// <summary> /// 参数 /// </summary> private string param = null; /// <summary> /// 语法 /// </summary> private string grammar = null; //错误消息通知托管 public delegate void delegdateOnerror(string Msg); private string path = null; /// <summary> /// 识别数据返回的事件 /// </summary> public event EventHandler<DataArrivedEventArgs> DataArrived; /// <summary> /// 识别过程结束的事件 /// </summary> public event EventHandler ISREnd; /// <summary> /// 正在识别 /// </summary> public event EventHandler Spoting; #endregion #region 构造函数 /// <summary> /// 构造函数,初始化引擎 /// </summary> /// <param name="appid">appid</param> /// <param name="param">参数</param> /// <param name="grammar">语法ID</param> /// <param name="path">路径</param> public MscNet(string appid, string param, string path, string grammar) { this.path = path; this.param = param; this.grammar = grammar; Re = new RecoErrors(); //引擎初始化,只需初始化一次 ret = MscDll.QISRInit("appid=" + appid); try { Re.GetError(ret); } catch (MscException ex) { RaiseError(ex.Message.ToString()); } } #endregion #region 公共方法 /// <summary> /// 开始识别语音 /// </summary> /// <param name="buffer">音频数据</param> public void InterpretVoice(byte[] buffer) { //用来指明用户本次识别的音频是否发送完毕 AudioStatus audioStatus = AudioStatus.ISR_AUDIO_SAMPLE_LAST; //端点检测(End-point detected)器所处的状态 EpStatus ep_status = EpStatus.ISR_EP_NULL; //识别器所处的状态 RecogStatus rec_status = RecogStatus.ISR_REC_NULL; //识别结果的状态 RecogStatus rslt_status = RecogStatus.ISR_REC_NULL; Loadgrammar(); int ret = MscDll.QISRAudioWrite(sess_id, buffer, (uint)buffer.Length, audioStatus, ref ep_status, ref rec_status); try { Re.GetError(ret); do { if (rslt_status == RecogStatus.ISR_REC_STATUS_INCOMPLETE) { Spoting(this, new EventArgs());//通知正在识别 } IntPtr p = MscDll.QISRGetResult(sess_id, ref rslt_status, 0, ref ret); Re.GetError(ret); if (p != IntPtr.Zero) { string tmp = PtrToStr(p); DataArrived(this, new DataArrivedEventArgs(tmp));//激发识别数据到达事件 } System.Threading.Thread.Sleep(500); } while (rslt_status != RecogStatus.ISR_REC_STATUS_SPEECH_COMPLETE); } catch (MscException ex) { RaiseError(ex.Message); } finally { try { ret = MscDll.QISRSessionEnd(sess_id, string.Empty); Re.GetError(ret); ISREnd(this, new EventArgs());//通知识别结束 } catch (MscException ex) { RaiseError(ex.Message); } } } /// <summary> /// 上传词库词库采用是utf8格式 /// </summary> /// <param name="txtFile">词库名称</param> /// <param name="path">词库路径</param> /// <param name="param">参数</param> /// <returns>返回词库编号</returns> public string GetExID(string txtFile, string path, string param) { string filePath = Path.Combine(path, txtFile); string tmp = string.Empty; if (!string.IsNullOrEmpty(filePath)) { try { sess_id = PtrToStr(MscDll.QISRSessionBegin(null, param, ref ret)); Re.GetError(ret); try { using (FileStream fs = File.Open(filePath, FileMode.Open)) { long len = fs.Length; byte[] buffer = new byte[len]; fs.Read(buffer, 0, (int)len); IntPtr p = MscDll.QISRUploadData(sess_id, "sces", buffer, (uint)fs.Length, "dtt=keylist", ref ret); Re.GetError(ret); if (p != IntPtr.Zero) { tmp = PtrToStr(p); } } } catch (FileNotFoundException ex) { RaiseError(ex.Message); } } catch (MscException ex) { RaiseError(ex.Message); } finally { try { ret = MscDll.QISRSessionEnd(sess_id, null); Re.GetError(ret); } catch (MscException ex) { RaiseError(ex.Message.ToString()); } } } else { RaiseError("路径不正确!"); } return tmp; } /// <summary> /// 对MSC的ISR部分进行逆初始化。 /// </summary> public void QISRFini() { try { ret = MscDll.QISRFini(); Re.GetError(ret); } catch (MscException ex) { RaiseError(ex.Message.ToString()); } } #endregion #region 受保护方法 /// <summary> /// 加载语法 /// </summary> private void Loadgrammar() { try { sess_id = PtrToStr(MscDll.QISRSessionBegin(grammar, param, ref ret)); Re.GetError(ret); } catch (MscException ex) { RaiseError(ex.Message); } } /// <summary> /// 指针转字符串 /// </summary> /// <param name="p">指向非托管代码字符串的指针</param> /// <returns>返回指针指向的字符串</returns> private string PtrToStr(IntPtr p) { List<byte> lb = new List<byte>(); try { while (Marshal.ReadByte(p) != 0) { lb.Add(Marshal.ReadByte(p)); p = p + 1; } } catch (AccessViolationException ex) { RaiseError(ex.Message); } return Encoding.Default.GetString(lb.ToArray()); } #endregion #region 事件 /// <summary> /// 错误通知事件 /// </summary> public event delegdateOnerror OnError; private void RaiseError(string Msg) { if (OnError != null) { OnError(Msg); } } /// <summary> /// 有识别数据返回的事件参数,包含了识别的文本结果 /// </summary> public class DataArrivedEventArgs : EventArgs { public string result; public DataArrivedEventArgs(string rs) { result = rs; } } #endregion
如果采用ABNF语法,只是与文本词库加载语法方式有点不一样:
/// <summary> /// 加载语法 /// </summary> private void Loadgrammar() { try { //sess_id = PtrToStr(MscDll.QISRSessionBegin(grammar, param, ref ret)); /*ABNF语法*/ sess_id = PtrToStr(MscDll.QISRSessionBegin(null, "rst=plain,sub=asr,ssm=1,aue=speex,auf=audio/L16;rate=16000,cfd=350", ref ret)); em.GetError(ret); string grammar1 = "#ABNF 1.0 GB2312;\n mode voice;\n language zh-cn;\n\n\n root $main;\n public $main = 我[想要]看$place1;\n $place1=足球;\n"; ret = MscDll.QISRGrammarActivate(sess_id, grammar1, "abnf", 0);//将语法ID传入QISRSessionBegin /*end */ em.GetError(ret); } catch (MscException ex) { RaiseError(ex.Message.ToString()); } }
常量的枚举:
#region 错误代号 enum ErrorCode { MSP_SUCCESS= 0, MSP_ERROR_FAIL = -1, MSP_ERROR_EXCEPTION = -2, /* General errors 10100(0x2774) */ MSP_ERROR_GENERAL = 10100, /* 0x2774 */ MSP_ERROR_OUT_OF_MEMORY = 10101, /* 0x2775 */ MSP_ERROR_FILE_NOT_FOUND = 10102, /* 0x2776 */ MSP_ERROR_NOT_SUPPORT = 10103, /* 0x2777 */ MSP_ERROR_NOT_IMPLEMENT = 10104, /* 0x2778 */ MSP_ERROR_ACCESS = 10105, /* 0x2779 */ MSP_ERROR_INVALID_PARA = 10106, /* 0x277A */ MSP_ERROR_INVALID_PARA_VALUE = 10107, /* 0x277B */ MSP_ERROR_INVALID_HANDLE = 10108, /* 0x277C */ MSP_ERROR_INVALID_DATA = 10109, /* 0x277D */ MSP_ERROR_NO_LICENSE = 10110, /* 0x277E */ MSP_ERROR_NOT_INIT = 10111, /* 0x277F */ MSP_ERROR_NULL_HANDLE = 10112, /* 0x2780 */ MSP_ERROR_OVERFLOW = 10113, /* 0x2781 */ MSP_ERROR_TIME_OUT = 10114, /* 0x2782 */ MSP_ERROR_OPEN_FILE = 10115, /* 0x2783 */ MSP_ERROR_NOT_FOUND = 10116, /* 0x2784 */ MSP_ERROR_NO_ENOUGH_BUFFER = 10117, /* 0x2785 */ MSP_ERROR_NO_DATA = 10118, /* 0x2786 */ MSP_ERROR_NO_MORE_DATA = 10119, /* 0x2787 */ MSP_ERROR_SKIPPED = 10120, /* 0x2788 */ MSP_ERROR_ALREADY_EXIST = 10121, /* 0x2789 */ MSP_ERROR_LOAD_MODULE = 10122, /* 0x278A */ MSP_ERROR_BUSY = 10123, /* 0x278B */ MSP_ERROR_INVALID_CONFIG = 10124, /* 0x278C */ MSP_ERROR_VERSION_CHECK = 10125, /* 0x278D */ MSP_ERROR_CANCELED = 10126, /* 0x278E */ MSP_ERROR_INVALID_MEDIA_TYPE = 10127, /* 0x278F */ MSP_ERROR_CONFIG_INITIALIZE = 10128, /* 0x2790 */ MSP_ERROR_CREATE_HANDLE = 10129, /* 0x2791 */ MSP_ERROR_CODING_LIB_NOT_LOAD = 10130, /* 0x2792 */ /* Error codes of network 10200(0x27D8)*/ MSP_ERROR_NET_GENERAL = 10200, /* 0x27D8 */ MSP_ERROR_NET_OPENSOCK = 10201, /* 0x27D9 */ /* Open socket */ MSP_ERROR_NET_CONNECTSOCK = 10202, /* 0x27DA */ /* Connect socket */ MSP_ERROR_NET_ACCEPTSOCK = 10203, /* 0x27DB */ /* Accept socket */ MSP_ERROR_NET_SENDSOCK = 10204, /* 0x27DC */ /* Send socket data */ MSP_ERROR_NET_RECVSOCK = 10205, /* 0x27DD */ /* Recv socket data */ MSP_ERROR_NET_INVALIDSOCK = 10206, /* 0x27DE */ /* Invalid socket handle */ MSP_ERROR_NET_BADADDRESS = 10207, /* 0x27EF */ /* Bad network address */ MSP_ERROR_NET_BINDSEQUENCE = 10208, /* 0x27E0 */ /* Bind after listen/connect */ MSP_ERROR_NET_NOTOPENSOCK = 10209, /* 0x27E1 */ /* Socket is not opened */ MSP_ERROR_NET_NOTBIND = 10210, /* 0x27E2 */ /* Socket is not bind to an address */ MSP_ERROR_NET_NOTLISTEN = 10211, /* 0x27E3 */ /* Socket is not listenning */ MSP_ERROR_NET_CONNECTCLOSE = 10212, /* 0x27E4 */ /* The other side of connection is closed */ MSP_ERROR_NET_NOTDGRAMSOCK = 10213, /* 0x27E5 */ /* The socket is not datagram type */ MSP_ERROR_NET_DNS=10214, /* Error codes of mssp message 10300(0x283C) */ MSP_ERROR_MSG_GENERAL = 10300, /* 0x283C */ MSP_ERROR_MSG_PARSE_ERROR = 10301, /* 0x283D */ MSP_ERROR_MSG_BUILD_ERROR = 10302, /* 0x283E */ MSP_ERROR_MSG_PARAM_ERROR = 10303, /* 0x283F */ MSP_ERROR_MSG_CONTENT_EMPTY = 10304, /* 0x2840 */ MSP_ERROR_MSG_INVALID_CONTENT_TYPE = 10305, /* 0x2841 */ MSP_ERROR_MSG_INVALID_CONTENT_LENGTH = 10306, /* 0x2842 */ MSP_ERROR_MSG_INVALID_CONTENT_ENCODE = 10307, /* 0x2843 */ MSP_ERROR_MSG_INVALID_KEY = 10308, /* 0x2844 */ MSP_ERROR_MSG_KEY_EMPTY = 10309, /* 0x2845 */ MSP_ERROR_MSG_SESSION_ID_EMPTY = 10310, /* 0x2846 */ MSP_ERROR_MSG_LOGIN_ID_EMPTY = 10311, /* 0x2847 */ MSP_ERROR_MSG_SYNC_ID_EMPTY = 10312, /* 0x2848 */ MSP_ERROR_MSG_APP_ID_EMPTY = 10313, /* 0x2849 */ MSP_ERROR_MSG_EXTERN_ID_EMPTY = 10314, /* 0x284A */ MSP_ERROR_MSG_INVALID_CMD = 10315, /* 0x284B */ MSP_ERROR_MSG_INVALID_SUBJECT = 10316, /* 0x284C */ MSP_ERROR_MSG_INVALID_VERSION = 10317, /* 0x284D */ MSP_ERROR_MSG_NO_CMD = 10318, /* 0x284E */ MSP_ERROR_MSG_NO_SUBJECT = 10319, /* 0x284F */ MSP_ERROR_MSG_NO_VERSION = 10320, /* 0x2850 */ MSP_ERROR_MSG_MSSP_EMPTY = 10321, /* 0x2851 */ MSP_ERROR_MSG_NEW_RESPONSE = 10322, /* 0x2852 */ MSP_ERROR_MSG_NEW_CONTENT = 10323, /* 0x2853 */ MSP_ERROR_MSG_INVALID_SESSION_ID = 10324, /* 0x2854 */ /* Error codes of DataBase 10400(0x28A0)*/ MSP_ERROR_DB_GENERAL = 10400, /* 0x28A0 */ MSP_ERROR_DB_EXCEPTION = 10401, /* 0x28A1 */ MSP_ERROR_DB_NO_RESULT = 10402, /* 0x28A2 */ MSP_ERROR_DB_INVALID_USER = 10403, /* 0x28A3 */ MSP_ERROR_DB_INVALID_PWD = 10404, /* 0x28A4 */ MSP_ERROR_DB_CONNECT = 10405, /* 0x28A5 */ MSP_ERROR_DB_INVALID_SQL = 10406, /* 0x28A6 */ MSP_ERROR_DB_INVALID_APPID = 10407, /* 0x28A7 */ /* Error codes of Resource 10500(0x2904)*/ MSP_ERROR_RES_GENERAL = 10500, /* 0x2904 */ MSP_ERROR_RES_LOAD = 10501, /* 0x2905 */ /* Load resource */ MSP_ERROR_RES_FREE = 10502, /* 0x2906 */ /* Free resource */ MSP_ERROR_RES_MISSING = 10503, /* 0x2907 */ /* Resource File Missing */ MSP_ERROR_RES_INVALID_NAME = 10504, /* 0x2908 */ /* Invalid resource file name */ MSP_ERROR_RES_INVALID_ID = 10505, /* 0x2909 */ /* Invalid resource ID */ MSP_ERROR_RES_INVALID_IMG = 10506, /* 0x290A */ /* Invalid resource image pointer */ MSP_ERROR_RES_WRITE = 10507, /* 0x290B */ /* Write read-only resource */ MSP_ERROR_RES_LEAK = 10508, /* 0x290C */ /* Resource leak out */ MSP_ERROR_RES_HEAD = 10509, /* 0x290D */ /* Resource head currupt */ MSP_ERROR_RES_DATA = 10510, /* 0x290E */ /* Resource data currupt */ MSP_ERROR_RES_SKIP = 10511, /* 0x290F */ /* Resource file skipped */ /* Error codes of TTS 10600(0x2968)*/ MSP_ERROR_TTS_GENERAL = 10600, /* 0x2968 */ MSP_ERROR_TTS_TEXTEND = 10601, /* 0x2969 */ /* Meet text end */ MSP_ERROR_TTS_TEXT_EMPTY = 10602, /* 0x296A */ /* no synth text */ /* Error codes of Recognizer 10700(0x29CC) */ MSP_ERROR_REC_GENERAL = 10700, /* 0x29CC */ MSP_ERROR_REC_INACTIVE = 10701, /* 0x29CD */ MSP_ERROR_REC_GRAMMAR_ERROR = 10702, /* 0x29CE */ MSP_ERROR_REC_NO_ACTIVE_GRAMMARS = 10703, /* 0x29CF */ MSP_ERROR_REC_DUPLICATE_GRAMMAR = 10704, /* 0x29D0 */ MSP_ERROR_REC_INVALID_MEDIA_TYPE = 10705, /* 0x29D1 */ MSP_ERROR_REC_INVALID_LANGUAGE = 10706, /* 0x29D2 */ MSP_ERROR_REC_URI_NOT_FOUND = 10707, /* 0x29D3 */ MSP_ERROR_REC_URI_TIMEOUT = 10708, /* 0x29D4 */ MSP_ERROR_REC_URI_FETCH_ERROR = 10709, /* 0x29D5 */ /* Error codes of Speech Detector 10800(0x2A30) */ MSP_ERROR_EP_GENERAL = 10800, /* 0x2A30 */ MSP_ERROR_EP_NO_SESSION_NAME = 10801, /* 0x2A31 */ MSP_ERROR_EP_INACTIVE = 10802, /* 0x2A32 */ MSP_ERROR_EP_INITIALIZED = 10803, /* 0x2A33 */ /* Error codes of TUV */ MSP_ERROR_TUV_GENERAL = 10900, /* 0x2A94 */ MSP_ERROR_TUV_GETHIDPARAM = 10901, /* 0x2A95 */ /* Get Busin Param huanid*/ MSP_ERROR_TUV_TOKEN = 10902, /* 0x2A96 */ /* Get Token */ MSP_ERROR_TUV_CFGFILE = 10903, /* 0x2A97 */ /* Open cfg file */ MSP_ERROR_TUV_RECV_CONTENT = 10904, /* 0x2A98 */ /* received content is error */ MSP_ERROR_TUV_VERFAIL = 10905, /* 0x2A99 */ /* Verify failure */ /* Error codes of IMTV */ MSP_ERROR_IMTV_SUCCESS = 11000, /* 0x2AF8 */ /* 成功 */ MSP_ERROR_IMTV_NO_LICENSE = 11001, /* 0x2AF9 */ /* 试用次数结束,用户需要付费 */ MSP_ERROR_IMTV_SESSIONID_INVALID = 11002, /* 0x2AFA */ /* SessionId失效,需要重新登录通行证 */ MSP_ERROR_IMTV_SESSIONID_ERROR = 11003, /* 0x2AFB */ /* SessionId为空,或者非法 */ MSP_ERROR_IMTV_UNLOGIN = 11004, /* 0x2AFC */ /* 未登录通行证 */ MSP_ERROR_IMTV_SYSTEM_ERROR = 11005, /* 0x2AFD */ /* 系统错误 */ /* Error codes of HCR */ MSP_ERROR_HCR_GENERAL = 11100, MSP_ERROR_HCR_RESOURCE_NOT_EXIST = 11101, /* Error codes of http 12000(0x2EE0) */ MSP_ERROR_HTTP_BASE = 12000, /* 0x2EE0 */ /*Error codes of ISV */ MSP_ERROR_ISV_NO_USER = 13000, /* 32C8 */ /* the user doesn't exist */ } #endregion #region ISR枚举常量 public enum AudioStatus { ISR_AUDIO_SAMPLE_INIT = 0x00, ISR_AUDIO_SAMPLE_FIRST = 0x01, ISR_AUDIO_SAMPLE_CONTINUE = 0x02, ISR_AUDIO_SAMPLE_LAST = 0x04, ISR_AUDIO_SAMPLE_SUPPRESSED = 0x08, ISR_AUDIO_SAMPLE_LOST = 0x10, ISR_AUDIO_SAMPLE_NEW_CHUNK = 0x20, ISR_AUDIO_SAMPLE_END_CHUNK = 0x40, ISR_AUDIO_SAMPLE_VALIDBITS = 0x7f /* to validate the value of sample->status */ } public enum EpStatus { ISR_EP_NULL = -1, ISR_EP_LOOKING_FOR_SPEECH = 0, ///还没有检测到音频的前端点 ISR_EP_IN_SPEECH = 1, ///已经检测到了音频前端点,正在进行正常的音频处理。 ISR_EP_AFTER_SPEECH = 3, ///检测到音频的后端点,后继的音频会被MSC忽略。 ISR_EP_TIMEOUT = 4, ///超时 ISR_EP_ERROR = 5, ///出现错误 ISR_EP_MAX_SPEECH = 6 ///音频过大 } public enum RecogStatus { ISR_REC_NULL = -1, ISR_REC_STATUS_SUCCESS = 0, ///识别成功,此时用户可以调用QISRGetResult来获取(部分)结果。 ISR_REC_STATUS_NO_MATCH = 1, ///识别结束,没有识别结果 ISR_REC_STATUS_INCOMPLETE = 2, ///正在识别中 ISR_REC_STATUS_NON_SPEECH_DETECTED = 3, ///保留 ISR_REC_STATUS_SPEECH_DETECTED = 4, ///发现有效音频 ISR_REC_STATUS_SPEECH_COMPLETE = 5, ///识别结束 ISR_REC_STATUS_MAX_CPU_TIME = 6, ///保留 ISR_REC_STATUS_MAX_SPEECH = 7, ///保留 ISR_REC_STATUS_STOPPED = 8, ///保留 ISR_REC_STATUS_REJECTED = 9, ///保留 ISR_REC_STATUS_NO_SPEECH_FOUND = 10 ///没有发现音频 } #endregion }
自定义异常:
[Serializable] //声明为可序列化的 因为要写入文件中 public class MscException : ApplicationException//由用户程序引发,用于派生自定义的异常类型 { /// <summary> /// 默认构造函数 /// </summary> public MscException() { } public MscException(string message) : base(message) { } public MscException(string message, Exception inner) : base(message, inner) { } } /// <summary> /// 是否出错 /// </summary> internal class RecoErrors { public RecoErrors() { } /// <summary> /// 是否发生错误. /// </summary> /// <param name="id">错误ID</param> public virtual void GetError(int id) { if (id != 0) { var ex = new MscException(((ErrorCode)id).ToString("G")); // var ex = new MscException(Enum.GetName(typeof(ErrorCode),id)); throw ex; } } }
讯飞语音支持边录边上传,不过我这里采用是一次性上传。起初我采用的是边录边上传,不过感觉有数字混合后识别正常率不好(还没跟讯飞那边沟通。),最后才使用一次性上传,毕竟语音文件也不是大就200KB一下。讯飞语音识别不支持多线程识别。
我做的这个语音识别产品,做成服务端与客户端。服务端:放在一个能连接外网机子上提供语音识别(做了一个简单队列),客户端:将音频数据采集后发送到局域网内的语音识别服务端进行识别。
以上有些代码是借助别人的,第一次写大家尽量不要吐槽,不过可以给点意见。大家相互学习…